mlsanityData to 14 Oct 2026

AntiPaSTO: Self-Supervised Honesty Steering via Anti-Parallel Representations

arXiv 2601.07473 ↗First posted 12 Jan 2026cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

As models grow more capable, humans cannot reliably verify what they say. Scalable steering requires methods that are internal, self-supervised, and transfer out-of-distribution; existing methods satisfy some but not all three. We introduce AntiPaSTO, which separates representations along an antiparallel axis (+1/-1 produce opposite shifts), with coherence constraints preventing collapse. Training uses only two contrasting words inserted into template sentences, with no preference labels. When we use 800 such synthetic pairs on Gemma-3-1B, AntiPaSTO beats prompting baselines by 6.9x Steering F1 on DailyDilemmas and wins on 5 of 6 tested value axes. We also find preliminary evidence that it maintains bidirectional control where prompting triggers refusal.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 27 of its 41 references that are in the map. First posted on arXiv 12 Jan 2026.

Builds on

27 of 41 references.

References · 27

  1. 2016Concrete Problems in AI Safetycited by 1,642
  2. 2023Inference-Time Intervention: Eliciting Truthful Answers from a Language Modelcited by 1,302
  3. 2023Steering Llama 2 via Contrastive Activation Additioncited by 1,085
  4. 2023Steering Language Models With Activation Engineeringcited by 922
  5. 2021Probing Classifiers: Promises, Shortcomings, and Advancescited by 876
  6. 2023A Simple and Effective Pruning Approach for Large Language Modelscited by 800
  7. 2022Discovering Latent Knowledge in Language Models Without Supervisioncited by 767
  8. 2023Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervisioncited by 467
  9. 2025AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoderscited by 218
  10. 2023Controlling Text-to-Image Diffusion by Orthogonal Finetuningcited by 208
  11. 2024Programming Refusal with Conditional Activation Steeringcited by 206
  12. 2022Goal Misgeneralization: Why Correct Specifications Aren't Enough For Correct Goalscited by 126
15 more
  1. 2021Goal Misgeneralization in Deep Reinforcement Learningcited by 123
  2. 2024Personalized Steering of Large Language Models: Versatile Steering Vectors Through Bi-directional Preference Optimizationcited by 118
  3. 2024Universal Neurons in GPT2 Language Modelscited by 102
  4. 2025Utility Engineering: Analyzing and Controlling Emergent Value Systems in AIscited by 67
  5. 2025Understanding (Un)Reliability of Steering Vectors in Language Modelscited by 47
  6. 2019Exposure Bias versus Self-Recovery: Are Distortions Really Incremental for Autoregressive Text Generation?cited by 35
  7. 2025HyperSteer: Activation Steering at Scale with Hypernetworkscited by 23
  8. 2025Adaptive Multi-Subspace Representation Steering for Attribute Alignment in Large Language Modelscited by 16
  9. 2025Steering the CensorShip: Uncovering Representation Vectors for LLM "Thought" Controlcited by 14
  10. 2025RepIt: Steering Language Models with Concept-Specific Refusal Vectorscited by 12
  11. 2025Rethinking Reward Model Evaluation Through the Lens of Reward Overoptimizationcited by 11
  12. 2025DeLoRA: Decoupling Angles and Strength in Low-rank Adaptationcited by 10
  13. 2025Distribution-Aligned Decoding for Efficient LLM Task Adaptationcited by 7
  14. 2025Deep Value Benchmark: Measuring Whether Models Generalize Deep Values or Shallow Preferencescited by 6
  15. 2025SSVD: Structured SVD for Parameter-Efficient Fine-Tuning and Benchmarking under Domain Shift in ASRcited by 4

Built on by

Not cited yet.